arXiv subjects

Chunlin Gong

Publications and source records attributed to Chunlin Gong.

11 recordsLinked to original sources

arXiv2026

CoT is Not the Chain of Truth: An Empirical Internal Analysis of Reasoning LLMs for Fake News Generation

From generating headlines to fabricating news, the Large Language Models (LLMs) are typically assessed by their final outputs, under the safety assumption that a refusal response signifies safe reasoning throughout the entire process. Challenging this assumption, our study reveals that during fake news generation, even when a model rejects a harmful request, its Chain-of-Thought (CoT) reasoning may still internally contain and propagate unsafe narratives. To analyze this phenomenon, we introduce a unified safety-analysis framework that systematically deconstructs CoT generation across model layers and evaluates the role of individual attention heads through Jacobian-based spectral metrics. Within this framework, we introduce three interpretable measures: stability, geometry, and energy to quantify how specific attention heads respond or embed deceptive reasoning patterns. Extensive experiments on multiple reasoning-oriented LLMs show that the generation risk rises significantly when the thinking mode is activated, where the critical routing decisions are concentrated in only a few contiguous mid-depth layers. By precisely identifying the attention heads responsible for this divergence, our work challenges the assumption that refusal implies safety and provides a new understanding perspective for mitigating latent reasoning risks.

cs.CL↗

arXiv2026

Manifold Learning with Implicit Physics Embedding for Reduced-Order Flow-Field Modeling

Nonlinear manifold learning (ML) based reduced-order models (ROMs) can substantially improve the quality of nonlinear flow-field modeling. However, noise and the lack of physical information often distort the dimensionality-reduction process, reducing the robustness and accuracy of flow-field prediction. To address this problem, we propose a novel manifold learning ROM with implicit physics embedding (IPE-ML). Starting from data-driven manifold coordinates, we incorporate physical parameters (e.g., angle of attack, Mach number) into manifold coordinates system by minimizing the prediction error of Gaussian process regression (GPR) model, thereby fine-tuning the manifold structure. These adjusted coordinates are then used to construct a flow-fields prediction model that predict nonlinear flow-field more accurately. The method is validated on two test cases: transonic flow-field modeling of the RAE2822 and supersonic flow-field modeling of the hexagon airfoil. The results indicate that the proposed IPE-ML can significantly improve the overall prediction accuracy of nonlinear flow fields. In transonic case, shock-related errors have been notably reduced, while in supersonic case the method can confine errors to small local regions. This study offers a new perspective on embedding physical information into nonlinear ROMs.

physics.flu-dyn↗

arXiv2025

OptFormer: Optical Flow-Guided Attention and Phase Space Reconstruction for SST Forecasting

Sea Surface Temperature (SST) prediction plays a vital role in climate modeling and disaster forecasting. However, it remains challenging due to its nonlinear spatiotemporal dynamics and extended prediction horizons. To address this, we propose OptFormer, a novel encoder-decoder model that integrates phase-space reconstruction with a motion-aware attention mechanism guided by optical flow. Unlike conventional attention, our approach leverages inter-frame motion cues to highlight relative changes in the spatial field, allowing the model to focus on dynamic regions and capture long-range temporal dependencies more effectively. Experiments on NOAA SST datasets across multiple spatial scales demonstrate that OptFormer achieves superior performance under a 1:1 training-to-prediction setting, significantly outperforming existing baselines in accuracy and robustness.

cs.CV↗

arXiv2025

CogniDir: Combating Cognitive Malicious Comments via Adaptive Distributional Learning for Robust Fake News Detection

The proliferation of Large Language Models (LLMs) has enabled a new class of psychologically grounded malicious comments, shifting fake news attacks from surface-level textual noise to deep cognitive and logical manipulation. This shift severely undermines existing detectors, which conventionally rely on static attack assumptions and fixed training distributions. To bridge this gap, we introduce CogniDir, an adaptive distributional learning framework that reformulates robust detection as a dynamic data mixture optimization problem for social media content safety. Grounded in cognitive psychology, we first formalize mechanism-specific cognitive adversarial paradigms to systematically expose deep-seated detector vulnerabilities. To address the vulnerability heterogeneity, CogniDir derives an information-theoretic score coupling empirical accuracy with probabilistic confidence, which is then mapped to adaptive sampling proportions through a Dirichlet-mean parameterization. This formulation enables smooth, feedback-driven reallocation of training exposure toward the most brittle attack mechanisms. Experimental results on three benchmarks demonstrate that CogniDir yields state-of-the-art robustness, improving F1 scores by up to 17.9% over competitive baselines under heterogeneous, AI-generated adversarial pressures.

cs.LG↗

arXiv2025

AFD-STA: Adaptive Filtering Denoising with Spatiotemporal Attention for Chaotic System Prediction

This paper presents AFD-STA Net, a neural framework integrating adaptive filtering and spatiotemporal dynamics learning for predicting high-dimensional chaotic systems governed by partial differential equations. The architecture combines: 1) An adaptive exponential smoothing module with position-aware decay coefficients for robust attractor reconstruction, 2) Parallel attention mechanisms capturing cross-temporal and spatial dependencies, 3) Dynamic gated fusion of multiscale features, and 4) Deep projection networks with dimension-scaling capabilities. Numerical experiments on nonlinear PDE systems demonstrate the model's effectiveness in maintaining prediction accuracy under both smooth and strongly chaotic regimes while exhibiting noise tolerance through adaptive filtering. Component ablation studies confirm critical contributions from each module, particularly highlighting the essential role of spatiotemporal attention in learning complex dynamical interactions. The framework shows promising potential for real-world applications requiring simultaneous handling of measurement uncertainties and high-dimensional nonlinear dynamics.

cs.LG↗

arXiv2025

STFM: A Spatio-Temporal Information Fusion Model Based on Phase Space Reconstruction for Sea Surface Temperature Prediction

The sea surface temperature (SST), a key environmental parameter, is crucial to optimizing production planning, making its accurate prediction a vital research topic. However, the inherent nonlinearity of the marine dynamic system presents significant challenges. Current forecasting methods mainly include physics-based numerical simulations and data-driven machine learning approaches. The former, while describing SST evolution through differential equations, suffers from high computational complexity and limited applicability, whereas the latter, despite its computational benefits, requires large datasets and faces interpretability challenges. This study presents a prediction framework based solely on data-driven techniques. Using phase space reconstruction, we construct initial-delay attractor pairs with a mathematical homeomorphism and design a Spatio-Temporal Fusion Mapping (STFM) to uncover their intrinsic connections. Unlike conventional models, our method captures SST dynamics efficiently through phase space reconstruction and achieves high prediction accuracy with minimal training data in comparative tests

cs.LG↗

arXiv2025

A Kernel Ridge Regression Combining Nonlinear ROMs for Accurate Flow Field Reconstruction with Discontinuities

Nonlinear reduced-order models (ROMs), represented by manifold learning (ML), can effectively improve the modeling accuracy of nonlinear flow fields with discontinuities. However, the inverse mapping from low-dimensional manifold coordinates to high-dimensional flow fields often introduces considerable reconstruction errors, leading to inaccuracy in the locations of discontinuities. To address this challenge, a novel reconstruction method is proposed to enhance the accuracy of reconstructing flow fields with discontinuities. The method employs kernel ridge regression (KRR) to construct a set of nonlinear modes rich in discontinuity information, sequentially these modes are nonlinearly combined with manifold coordinates to achieve accurate flow field reconstruction. The proposed reconstruction method is validated to reconstruct the transonic flow fields over RAE2822 airfoil. Comparison results demonstrate that the method achieves superior reconstruction accuracy compared to existing approaches, especially in reconstructing flow fields' discontinuous regions and precisely capturing discontinuities. This work provides an effective and highly interpretable solution for improving the accuracy of nonlinear ROMs in discontinuous flow fields modeling.

physics.flu-dyn↗

arXiv2024

Unsteady aerodynamic prediction using limited samples based on transfer learning

In this study, a method for predicting unsteady aerodynamic forces under different initial conditions using a limited number of samples based on transfer learning is proposed, aiming to avoid the need for large-scale high-fidelity aerodynamic simulations. First, a large number of training samples are acquired through high-fidelity simulation under the initial condition for the baseline, followed by the establishment of a pre-trained network as the source model using a long short-term memory (LSTM) network. When unsteady aerodynamic forces are predicted under the new initial conditions, a limited number of training samples are collected by high-fidelity simulations. Then, the parameters of the source model are transferred to the new prediction model, which is further fine-tuned and trained with limited samples. The new prediction model can be used to predict the unsteady aerodynamic forces of the entire process under the new initial conditions. The proposed method is validated by predicting the aerodynamic forces of free flight of a high-spinning projectile with a large extension of initial angular velocity and pitch angle. The results indicatethat the proposed method can predict unsteady aerodynamic forces under different initial conditions using 1/3 of the sample size of the source model. Compared with direct modeling using the LSTM networks, the proposed method shows improved accuracy and efficiency.

physics.flu-dyn↗

arXiv2023

An accuracy-enhanced transonic flow prediction method fusing deep learning and reduced-order model

It's difficult to accurately predict the flow with shock waves over an aircraft due to the flow's strongly nonlinear characteristics. In this study, we propose an accuracy-enhanced flow prediction method that fuses deep learning and reduced-order model to achieve fast flow field prediction for various aerodynamic shapes. First, we establish the convolutional neural network-proper orthogonal decomposition (CNN-POD) model for mapping geometries to the entire flow field. Next, local flow regions containing nonlinear flow structures are identified through POD reconstruction for enhanced modeling. Then, a new CNN model is employed to map geometries to the local flow field. The proposed method is finally applied in predicting transonic flow over airfoils. The results indicate that the proposed enhanced DNN method can reduce the prediction error of flow properties, particularly in the regions with shock waves (up to 13%-46%). Additionally, the better efficiency and robustness of the proposed methods have been validated in comparison to existing methods.

physics.flu-dyn↗

arXiv2018

Global optimization of expensive black-box models based on asynchronous hybrid-criterion with interval reduction

In this paper, a new sequential surrogate-based optimization (SSBO) algorithm is developed, which aims to improve the global search ability and local search efficiency for the global optimization of expensive black-box models. The proposed method involves three basic sub-criteria to infill new samples asynchronously to balance the global exploration and local exploitation. First, to capture the promising possible global optimal region, searching for the global optimum with genetic algorithm (GA) based on the current surrogate models of the objective and constraint functions. Second, to infill samples in the region with sparse samples to improve the global accuracy of the surrogate models, a grid searching with Latin hypercube sampling (LHS) with the current surrogate model is adopted to explore the sample space. Third, to accelerate the local searching efficiency, searching for a local optimum with sequential quadratic programming (SQP) based on the local surrogate models in the reduced interval, which involves some samples near the current optimum. When the new sample is too close to the existing ones, the new sample should be abandoned, due to the poor additional information. According to the three sub-criteria, the new samples are placed in the regions which have not been fully explored and includes the possible global optimum point. When a possible global optimum point is found, the local searching sub-criterion captures the local optimum around it rapidly. Numerical and engineering examples are used to verify the efficiency of the proposed method. The statistical results show that the proposed method has good global searching ability and efficiency.

stat.ML↗

arXiv2017

A sequential surrogate method for reliability analysis based on radial basis function

A radial basis function (RBF) based sequential surrogate reliability method (SSRM) is proposed, in which a special optimization problem is solved to update the surrogate model of the limit state function (LSF) iteratively. The objective of the optimization problem is to find a new point to maximize the probability density function (PDF), subject to the constraints that the new point is on the approximated LSF and the minimum distance to the existing points is greater than or equal to the given distance. By updating the surrogate model with the new points, the surrogate model of the LSF becomes more and more accurate in the important region with a high failure probability and on the LSF boundary. Moreover, the accuracy of the unimportant region is also improved within the iteration due to the minimum distance constraint. SSRM takes advantage of the information of PDF and LSF to capture the failure features, which decreases the number of the expensive LSF evaluations. Six numerical examples show that SSRM improves the accuracy of the surrogate model in the important region around the failure boundary with small number of samples and has better adaptability to the nonlinear LSF, hence increases the accuracy and efficiency of the reliability analysis.

stat.CO↗